Supported models for live agents¶

Supported in ADKPython v0.1.0

Live agents require a model that can hold a bidirectional connection; a standard Gemini model will not. For the models ADK supports outside live agents, and for non-Gemini providers, see Models for agents.

Live models¶

Live agents run on models that take audio in and produce audio out, end to end, with no intermediate text-to-speech stage. That is what gives them human-like speech with natural prosody, and it is what a standard Gemini model cannot do over a bidirectional connection.

Model	AI Studio	Agent Platform
Gemini 2.5 Flash Live	`gemini-2.5-flash-native-audio-preview-12-2025` (Preview)	`gemini-live-2.5-flash-native-audio` (GA)
Gemini 3.1 Flash Live	`gemini-3.1-flash-live-preview` (Preview)	Not available

Gemini 2.5 Flash Live is one model with a different ID on each backend; the features are the same either way. gemini-live-2.5-flash-native-audio is ADK's LlmAgent.DEFAULT_LIVE_MODEL, the only Live model that is publicly available, and the model used in this section's examples.

Gemini 3.1 Flash Live is the newer model and is lower latency, but it is AI Studio only and it drops features that 2.5 has see Per-model feature support before you switch.

Choosing a backend¶

Live models are reached through one of two backends. ADK talks to both with the same code; you switch with environment variables, so you can develop on one and deploy on the other.

	AI Studio	Agent Platform
Full name	Google AI Studio	Gemini Enterprise Agent Platform
Best for	Prototyping, development	Production, enterprise
Auth	API key (`GOOGLE_API_KEY`)	Cloud credentials (`GOOGLE_CLOUD_PROJECT`, `GOOGLE_CLOUD_LOCATION`)
Setup	API key only	Cloud project setup
Limits	Session duration and concurrency	Session duration and concurrency

Switch with the GOOGLE_GENAI_USE_ENTERPRISE environment variable (FALSE for AI Studio, TRUE for Agent Platform); no code changes. See the quickstarts for setup.

Agent Platform: the global location is not supported

Live models are not available at GOOGLE_CLOUD_LOCATION=global. Use a regional endpoint such as us-central1, us-east1, or asia-northeast1, and check it against the endpoint-locations table in Agent Platform locations before deploying.

These models produce audio directly, with natural prosody, and detect the conversation language on their own. What you configure on top voices, transcription, turn detection is described in Configuration.

One property is fixed at the model level: Live models produce audio only. They do not support the TEXT response modality, so to get text alongside speech you use audio transcription.

Per-model feature support¶

A few RunConfig and tool settings depend on which model you are running:

Feature	Gemini 2.5 Flash Live	Gemini 3.1 Flash Live
Proactivity and affective dialog	Opt-in via `RunConfig`	Not supported
`response_scheduling` on tools	Supported	Not supported; function calling is synchronous, so the model stays silent until you return the tool response
Thinking control	`thinking_budget`	`thinking_level` (`minimal`, `low`, `medium`, `high`)

Moving from 2.5 to 3.1

Leaving RunConfig.proactivity or RunConfig.enable_affective_dialog set is the most common upgrade failure remove them. Two more differences bite client code: a single server event can now carry several content parts at once, so iterate over event.content.parts instead of reading parts[0]; and turn coverage now defaults to including all detected audio activity and video frames, which changes token costs if you stream video continuously. See the upstream migration notes.

Platform limits and quotas¶

Both backends cap how long a connection and a session can run and how many sessions run at once. These numbers change, so treat the upstream documentation as authoritative and verify before you rely on a limit in production.

Limit	AI Studio	Agent Platform
Session duration, audio-only	15 min	15 min
Session duration, audio + video	2 min	2 min
Connection lifetime	~10 min	~10 min
Concurrent sessions	See rate limits	Up to 1,000 per project on pay-as-you-go; no limit with Provisioned Throughput

Agent Platform additionally caps a conversation session at 10 minutes by default, separately from the audio-only limit above.

Enabling context window compression lets a session be extended past the duration limits. On Agent Platform, request concurrent-session increases from the Cloud Console Quotas page under "Bidi generate content concurrent requests". Verify the current numbers against the AI Studio, Gemini API rate limits, and Agent Platform documentation.

How to handle model names¶

Read the model name from an environment variable rather than hard-coding it. The same model has a different ID on AI Studio and Agent Platform, so an .env var is what lets one codebase target both backends, and it insulates you from model deprecations.

Recommended Pattern:

import os
from google.adk.agents import Agent

# Use environment variable with fallback to a sensible default
agent = Agent(
    name="my_agent",
    model=os.getenv("DEMO_AGENT_MODEL", "gemini-live-2.5-flash-native-audio"),
    tools=[...],
    instruction="..."
)

Why use environment variables:

Backend-specific IDs: The same model is named differently on AI Studio and Agent Platform, so moving between them means changing the model ID. An env var keeps that out of your code
Model availability changes: Models are released and deprecated regularly. A live agent written a year ago should not be pinned in code to a model that no longer exists
Environment-specific configuration: Use different models for development, staging, and production

Configuration in .env file:

# AI Studio
DEMO_AGENT_MODEL=gemini-2.5-flash-native-audio-preview-12-2025

# AI Studio, if you do not need proactivity, affective dialog, or non-blocking tools
# DEMO_AGENT_MODEL=gemini-3.1-flash-live-preview

# Agent Platform
# DEMO_AGENT_MODEL=gemini-live-2.5-flash-native-audio

Environment Variable Loading Order

When using .env files with python-dotenv, you must call load_dotenv() before importing any modules that read environment variables. Otherwise, os.getenv() will return None and fall back to the default value, ignoring your .env configuration.

Correct order in main.py:

from dotenv import load_dotenv
from pathlib import Path

# Load .env file BEFORE importing agent
load_dotenv(Path(__file__).parent / ".env")

# Now safe to import modules that use environment variables
from google_search_agent.agent import agent

Incorrect order (will not work):

from dotenv import load_dotenv
from google_search_agent.agent import agent  # Agent reads env var here

# Too late! Agent already initialized with default model
load_dotenv(Path(__file__).parent / ".env")

This is a Python import behavior: when you import a module, its top-level code executes immediately. If your agent module calls os.getenv("DEMO_AGENT_MODEL") at import time, the .env file must already be loaded.

Selecting the right model:

Choose a backend: AI Studio for prototyping, Agent Platform for production. This picks the ID column in the table above, and on Agent Platform it settles the model too Gemini 2.5 Flash Live is the only Live model there
Check current availability: Refer to the model table above and the official documentation
Configure environment variable: Set the model name in your .env file and read it from there when constructing the agent

Model compatibility and availability¶

For the latest information on model compatibility and availability:

AI Studio: See the Gemini models documentation and the Live API capabilities guide
Agent Platform: See the Live API overview and the Agent Platform model documentation

Always verify model availability and feature support in the official documentation before deploying to production.